January_AMP_Digital

A D V A N C E D M A T E R I A L S & P R O C E S S E S | J A N U A R Y 2 0 1 8 2 4 For example, carburizing opera- tions have been modeled for several decades, ranging from analytical solu- tions of Fick’s diffusion equations to numerical solutions. In one case [1,2] , us- ing a cost modeling approach, having thermal, process, and numerical solu- tions of the diffusion process, param- eters such as set-point temperatures and time of boost and diffusion stages were optimized. This resulted in quan- titatively defining four distinct process regimes of (a) rework with incomplete carburizing, (b) rejection due to grain growth, (c) suboptimal regime of pro- ductivity and energy use, and (d) a nar- row optimal regime. Deployment of this modeling approach resulted in a 14% energy reduction and 20% productiv- ity enhancement for a battery of seal- quenched carburizing furnaces [2] . Such a deterministic modeling approach provides excellent insights for under- standing the phase transformation and distortion mechanisms [3] for specific components and process cycles. On the other hand, probabilistic approaches [4] are needed to understand batch-to-batch variations in chemistry and process parameters and their im- pact on component distortion during carburizing operations. Further, in re- cent work [5] , the effectiveness of analyt- ics techniques in leveraging significant materials, manufacturing, and quality data in conjunction with deterministic diffusion models was demonstrated in a modern heat treating shop with 12 carburizing furnaces. In this study, prin- cipal component analysis was used to reduce the multidimensional data and find key input variables impacting qual- ity. In this work, neural network models were used to predict quality parame- ters based on input variables. These da- ta-based models were synergized with diffusion models to determine the op- erating regimes along with rationaliza- tion of recipes as well as furnace health triggers. This significant analytics work demonstrated the potential of a 12.5% productivity increase [5] along with en- ergy reduction, ease of operation, and carbon footprint reduction (Fig. 2). The power of analytics was also demonstrated in a highly automat- ed batch annealing operation, where around 120 tons of cold rolled steel coils were annealed. In this work [6] , pro- cess and quality data was analyzed and it was demonstrated through analyt- ics techniques that the inherent con- trol system did not accurately capture the heat transfer and resultant phase transformation in the coil. By creating an accurate transfer function, this error was rectified and a potential 9% pro- ductivity enhancement in this annual one-million ton, cold rolling operation was demonstrated. In another batch annealing operation of tin-plated steel coils, the analytics approach resulted in TABLE 1 — TECHNOLOGY TRENDS AND MATERIALS ENGINEERING Technology Trend Materials Engineering Context Maturity Data analytics Optimizing materials processing for productivity, efficiency, energy, cost Mature, in use; more intentional use needed Machine learning and deep learning Unravelling patterns of decision-making in terms of materials, process, or recipe selection Mature and can be leveraged Digital thread, digital twin Optimizing and tracking component and product development across value chain Infrastructure not ready; use cases being formulated Blockchain Tracking genuineness and process compliance Very early Industry 4.0 or IoT Recipe management, predictive asset and process management, process optimization Infrastructure getting ready, use cases being formulated Autonomous vehicles Indirect implications due to design and safety expectations Very early Electric vehicles Materials palette impact, supply chain considerations, changes in the basis of materials selection and manufacturing Rapidly maturing; needs urgent attention Fig. 2 — Data analytics approach for a manufacturing operation.